Tutorial 5 intro: AUVLib – Sonar data processing for machine learning

Tutorial 5 intro: AUVLib – Sonar data processing for machine learning

In underwater surveys, many different kinds of sonars are used to capture different characteristics of the seafloor. In industrial surveys in particular, large amounts of data are captured and saved for offline processing. Using commercial software, the data can be inspected visually, and processed to produce stitched mosaics or bathymetric maps. In the new era of machine learning, new possibilities open up to further exploit the large amounts of data produced during underwater surveys. In this tutorial, we will introduce the participants to auvlib, an open-source python-based library that facilitates the process of applying machine learning to underwater sonar data. auvlib is able to read many common sonar file formats, and can achieve common tasks such as filtering data, producing bathymetric maps, and visualizing the data. In addition, we will show how to drape bathymetric maps with sidescan data in order to produce correspondences between the sensors that can be used for deep learning. The participants will thus use auvlib to make the full journey from raw sonar data to a trained neural network.

__

The tutorials are consisted of introduction presentation of 30 minutes and hands-on part of 60min. Tutorial hands-on and demos are divided in three groups which will rotate together every hour: